Most proteins for medical, cosmetic or industrial applications are produced as recombinant proteins in cultivated microbial or eukaryotic cells. For that a gene encoding the protein of interest is inserted into the organism/cells of choice, the organism/cells carrying said gene is cultivated in a medium comprising all essential nutrients to allow for growth and expression of said gene, resulting in the production of the protein of interest. If the protein of interest is secreted by the cells into the medium, the cells and the medium are separated from each other using centrifugation or filter membranes. The recovered cell-free protein containing medium is then further processed through purification steps to remove host cell proteins, DNA and other contaminants.
In general, there are two different production alternatives for the harvest of recombinantly produced proteins, continuous and batch harvest. When applying continuous harvest, the cultivation media is continuously slowly removed from the cell cultivation vessel during the production phase and fresh medium is simultaneously added. Continuous harvest is chosen when the cells are slowly growing and/or the process facilitates a high cell concentration or if the product must be removed fast from the cultivation to protect it from degradation. Batch harvesting is performed at one defined point, where the cells are removed in one step and thereafter they are normally discarded. It is technically easier to run a batch harvesting compared to a continuous one, but the optimal harvest procedure must be determined in each specific case depending on product and cell type. In some cases, where the product is not secreted, the cell membrane must be destroyed in order to recover the product. It is, however, preferred to keep the cell membrane intact, if possible, in order to avoid contamination of the product with DNA and host cell proteins. Continuous harvest normally gives a higher total productivity compared to batch harvest as the cells can produce for a longer period of time. To minimize the contamination of the product, it is preferable to choose the harvest method which releases the lowest amount of DNA and host cell proteins. In a special mode of batch harvesting, where the cell membranes are kept intact, a considerable improvement in productivity can be achieved, if the cell can be reused. This cyclic batch harvest method is particularly applicable for slowly growing (valuable) production cells.
To obtain high yields of the protein, it is important to optimize the process to achieve a high productivity of the product. The main efforts so far to improve the yield of a recombinantly produced protein have focused on the molecular insert (vectors, enhancers, promoters, etc.) to optimize the expression system, the conditions under which the cells are cultured and the actual purification steps. For example stabilizers such as protease inhibitors are added into the medium and the purification procedures are set up in the presence of protease inhibitors to reduce the loss of the product protein. However, it is often very difficult to find a protease inhibitor which can be used during cultivation, as protease inhibitors also tend to inhibit cell growth and protein production. As soon as the cells have been removed, it is easier to find a suitable protease inhibitor. This is described, for example, in U.S. Pat. No. 5,831,026, where EDTA is added to inhibit metallo-proteases. Moreover, it is to be noted that any stabilizing agents added need to be removed again at some point of the production process to obtain a pure recombinant protein product. Thus there is still a need to improve the existing methods of recombinant protein production to gain higher yields of the recombinant protein.
Another problem encountered especially when utilizing mammalian cells as production hosts is that the secretion of the produced proteins is rather low. It is apparent that secreted products often adhere to the cell membrane and that this has an influence on the product release. In some cases, the retardation can be inhibited by physiological conditions (i.e. the environment in which the cells are cultivated), whereas in some cases non-physiological conditions must be applied. The total disruption of cells should be avoided, if possible, as this releases DNA and host cell proteins, which need to be removed later in the purification process.
It is known in the art that an increase in salt concentrations (for instance, NaCl), accompanied by the addition of detergent and/or by adjusting a specific pH can in some cases release the bound proteins. For instance, K. Berman et al., Mol. Cell Biol. Res. Com. 4, 337-344 (2001) emphasizes that in the production of p38 homologs in HEK293 cells, the cells could be stimulated with sorbitol or 0.7 M NaCl for 10 minutes prior to harvest. A. B. Vaandrager et al., J. Biol. Chem., Vol. 271, No. 12, pp. 7025-7029 (1996) discloses that the cultivation of HEK293 cells expressing rat cGKII resulted in recovery of 90-95% of the expressed cGKII, and that the enzyme could be released from membranes by a combination of detergent (1% Triton® X-100) and high salt (0.5 M NaCl) but not by detergent or high salt alone. A. Denys et al., Biochem J., 336, 689-697 (1998) disclose that a protein was released from human T-lymphocyte cells using a washing procedure including 0.6 M NaCl. However, not all (70%) protein could be released even if NaCl concentration was raised to 1 M. Moreover, low pH, such as pH 4 did not release all of the bound protein (34%), while a combination of low pH and increased salt concentration (0.5 M NaCl, 0.2 M glycine, pH 4) released all bound protein. G. Grass et al., Infection and Immunity, p. 213-228 (2004) discloses that in the bacterial expression of metalloproteinases, the washing procedure with high ionic strength buffer (3 M NaCl) did not release the protein. The protein could, however be released by butanol or a detergent. C. M. Mounier et al., J. Biol. Chem. 279, No. 24, pp. 25024-25038 (2004) reports for HEK293 and CHO cells that proteins expressed by said cells bind to the cell surface. This could be inhibited by increased salt concentration (NaCl) in the range of 0.12 to 1 M. At 1 M all proteins seemed to be released. In one example with HEK293 the cells were treated with 1M NaCl and the released protein increased three times. M. Fannon et al., Biochemistry 39, 1434-1445 (2000) reports on the binding of proteins to fibroblasts. Three different washing procedures were compared, high salt (2 M NaCl, pH 7.4), low pH (20 mM sodium acetate, pH 4) and high salt, low pH (2 M NaCl, pH 4). All buffers do under certain circumstances release the protein. High salt and low pH is effective in all experiments. M. E. Zuber et al., J. Cell Physiology, 170:217-227 (1997) reports that the protein binding to the surface of CHO cells could be inhibited by high salt/low pH treatment (2 M NaCl, pH 4). J. Norbeck et al., FEMS Microbiology Letters 137, p. 1-8 (1996) reports about yeast cells which were subjected for 0.4 M NaCl for a period of 1.5 h during growing and about the effects thereof on the expression rate of various proteins. Finally, P. M. Dey et al., Planta 202:179-187 reports on the isolation of hydroxyproline-rich glycopoteins from suspension-cultured potato cells by washing the potato cells with a solution containing 50 mM CaCl2 and 2 mM ascorbic acid.
In view of the above it is apparent that a general method to increase the recovery of recombinant proteins in eukaryotic or mammalian expression systems is still desirable. Of particular interest are methods for the serum-free production of proteins which are needed for medical applications (such as plasma proteins including blood clotting factors which are required for the treatment of hemophilic disorders) and for which the serum-free product is desirable for obvious reasons. Hemophiliacs are suffering from hemorrhagic morbidity caused by the disturbed function of protein components of the blood coagulation cascade. Depending on the affected clotting factor, the hemophilia is classified in two types, hemophilia A and B, in both of which the conversion of soluble fibrinogen to an insoluble fibrin-clot is inhibited. They are recessive X-chromosomally-linked genetic disorders affecting mainly the male population.
Hemophilia A affects 1-2 individuals per 10.000 males. This is a genetic disorder that affects the ability of the blood to form an effective clot and thereby results in prolonged bleeding. As hemophilia A is an X-chromosome linked recessive disorder, almost exclusively men are affected. It is caused by the deficiency or absence of factor VIII, a very large glycoprotein (Mr approximately 330 kDa (Furie B., Furie B. C., Cell (1988) 53, 505-518; the sequence thereof is given in SEQ ID NO:2)), which represents an important element of the blood coagulation cascade. The polypeptide sequence of FVIII can be subdivided in three regions, an N-terminal region consisting of the so-called A1 and A2-domains, a central B-domain region and a C-terminal region composed of the A3, C1 and C2 domains. In the blood coagulation factor VIII occurs as an inactive precursor. It is bound tightly and non-covalently to von Willebrand Factor (vWF), which acts as a stabilizing carrier protein. Proteolytical cleavage of factor VIII by thrombin at three specific positions (740, 1372, 1689; see SEQ ID NO:2) leads to its dissociation from vWF and releases the procoagulant function within the cascade. In its active form factor VIII functions as a cofactor for factor IXa, thereby accelerating the proteolytic activation of factor X by several orders of magnitude.
Hemophilia B occurs in about 1 of 25,000 males. It is characterized by a deficiency of the serine protease factor IX (Christmas factor; see SEQ ID NO:11). The gene encoding factor IX is localized on the X-chromosome (locus Xq27) making hemophilia B an X-chromosome linked recessive disorder. This 415 amino-acid polypeptide is synthesized in the liver as a 56 kDa glycoprotein. In order to attain its proper function a posttranslational carboxylation step is required which only occurs in the presence of vitamin K.
Treatment of these types of bleeding disorders traditionally involves infusions of human plasma-derived protein concentrates of the missing factor(s), i.e. replacement therapy. Although this method represents an efficient therapy for hemophiliacs, it carries the risk of transmission of various infectious agents, such as viruses causing hepatitis or AIDS, or thromboembolic factors. Alternatively several recombinant DNA techniques for the production of clotting factors have been described. The corresponding cDNAs of wild type factor VIII and factor IX have been isolated and cloned into suitable expression vectors (EP-A-160457; WO-A-86/01961, U.S. Pat. Nos. 4,770,999, 5,521,070 and 5,521,070).
In the case of factor VIII, recombinant expression of subunits for the production of complexes showing coagulant activity is known in the art (e.g., from EP-A-150735, EP-A-232112, EP-A-0500734, WO-91/07490, WO-95/13300 U.S. Pat. Nos. 5,045,455 and 5,789,203). Moreover, the expression of truncated cDNA-versions partially or entirely lacking the sequence coding for the highly glycosylated B-domain have been described (e.g. in WO-86/06101, WO-87/04187, WO-87/07144, WO-88/00381, EP-A-251843, EP-A-253455, EP-A-254076, U.S. Pat. Nos. 4,868,112 and 4,980,456, EP-A-294910, EP-A-265778, EP-A-303540, WO-91/09122 and WO 01/70968.
The following passages provide details on human factor VIII because it was chosen as a model recombinant protein to illustrate the present invention.
The gene encoding the factor VIII protein is situated at the tip of the long arm of the X chromosome on locus Xq28. It spans over 186 kb, and thus is one of the largest genes known. The factor VIII gene comprises 26 exons and its transcription and subsequent processing results in a 9-kb mRNA. Translation of this mRNA leads to a polypeptide chain of 2351 amino acids, containing a signal peptide of 19 and a mature protein of 2332 amino acids (see SEQ ID NOs:1 and 2). Analysis of the primary structure determined from the cloned factor VIII cDNA revealed the organization in structural domains occurring in the order A1-a1-A2-a2-B-a3-A3-C1-C2.
The short spacers a1, a2 and a3 are so-called acidic regions containing clusters of Asp and Glu residues and are in literature often included in the A-domains resulting in the slightly simplified domain structure A1-A2-B-A3-C1-C2. Following translation and posttranslational modification, the primary translation product, having a molecular mass of approximately 300 kDa, undergoes intracellular proteolysis when leaving the Golgi apparatus processing the primary translation product into an amino terminal heavy chain of 90-210 kDa (A1-a1-A2-a2-B) and a carboxy terminal light chain of 80 kDa (a3-A3-C1-C2), giving rise to the heterodimeric molecule circulating in blood plasma. In this heterodimeric molecule the heavy and light chain of factor VIII are noncovalently linked by divalent metal ions. The span in molecular weights of the heavy chain is the result of different degrees of proteolytic cleavage of the B-domain. The more or less truncated B-domain remains attached to the A2-domain. The B-domain does not seem to have an influence on the biological activity of the FVIII molecule. This is supported by the fact that during activation of the FVIII the entire B-domain is cleaved off. Immediately after its release into the bloodstream, the FVIII heterodimer interacts with a carrier protein called “von Willebrand factor” (vWF). This interaction stabilizes the heterodimeric structure of FVIII increasing the half-life of FVIII in the blood circulation. Furthermore the complex-formation with vWF prevents the premature binding of factor VIII to cell membranes and components of the tenase complex. Also proteolytic cleavage of the FVIII molecule is to some extent prevented with the molecule being non-covalently bound to vWF. However, thrombin cleavage of FVIII is still possible and results in a loss of affinity to vWF and the conversion of FVIII to its active form.
As could be seen in the preceeding paragraph, factor VIII is a complex glycoprotein resulting in a difficult production process to maintain structural integrity and stability of the protein. Especially the B-domain harboring totally 19 of the altogether 25 N-linked glycosylation sites makes manufacturing of the full length protein difficult, as incorrect glycosylation always bears the risk of immunogenic reactions against the product. The function of the B-domain is not completely elucidated yet, but it has been found that this domain is not essential for the haemostatic function of factor VIII (Sandberg et al., Seminars in Hematology 38, Suppl 4, 24-31 (2001). This observation has been made both in vitro and in vivo for human plasma-derived factor VIII that lacks the entire B-domain, as well as for multiple forms of recombinant factor VIII molecules lacking the entire B-domain. Plasma-derived B-domain deleted factor VIII can be purified from plasma-derived factor VIII concentrates as these concentrates contain multiple active forms of factor VIII ranging in size from 170 kDa to 280 kDa most likely resulting from differences in the length of the B-domain still contained in the heterodimeric protein, supporting the finding that the B-domain is not essential for the biological activity of factor VIII (Eriksson et al., Seminars in Hematology 38, Suppl. 4, 24-31 (2001). In addition to its increased structural stability, transfection of eukaryotic cells with cDNA of the B-domain deleted factor VIII also yielded improved expression levels of the protein (Herlitschka et al., Journal of Biotechnology 61, 165-173 (1998)). These features resulted in one B-domain deleted recombinant factor VIII product being available on the market, showing comparable safety and efficacy as full length recombinant and plasma-derived factor VIII.
Generally deletion of the B-domain has been done on the cDNA-level resulting in the reduction of the overall size of the full-length factor VIII molecule by approximately 40% from 2332 amino acids to 1440 amino acids. The C-terminus of the heavy chain and the N-terminus of the light chain has in some cases been joined using a short amino acid linker replacing the entire B-domain with its 908 amino acids such in WO 00/49147 and WO 01/70968). The N-terminus of the linker described in these references was derived from the N-terminus of the B-domain whereas the C-terminus consists of a specially designed linker sequence. Like the full length recombinant factor VIII and the plasma-derived factor VIII, the majority of the B-domain deleted factor VIII is secreted as a non-covalently linked heterodimer of the heavy and the light chain. Also a small amount of non-cleaved single chain B-domain deleted recombinant factor VIII with a molecular weight of 170 kDa is secreted. Extensive studies have shown that binding to von Willebrand factor and activation by thrombin cleavage as well as interaction with several other physiologically relevant molecules is comparable to that described for the natural human factor VIII.
Recombinant factor VIII products exist. As the abundance of mRNA transcripts is very low (only 10−5 of the total mRNA of the liver) it took long to obtain the complete cDNA transcript of the protein. With this major breakthrough in the 1980s and the successful transfection of CHO cells with the cDNA, the first recombinant factor VIII product was introduced to the market in 1992. Since then, the annual sales of recombinant factor VIII preparations have reached values >1 billion US$ (Schmid: Pocket Guide to biotechnology and genetic engineering; Wiley-VCH (2003)). Currently four different recombinant factor VIII preparations are available on the market. The manufacturers of these four preparations cover approximately 60% of the demand for factor VIII preparations in the developed countries. However, capacity is still not sufficient and methods to increase the yield of a production process for a recombinant protein will be particularly beneficial for the production of recombinant factor VIII.
In the production of a recombinant B-domain deleted FVIII according to WO 01/70968, HEK cells were transformed with a gene for FVIII, transformed cells were cultivated and FVIII was secreted. During the harvest of the cells, the FVIII molecule and the cells were separated using centrifugation or filter membranes. The recovered cell-free FVIII containing media was then further processed through purification steps to remove host cell proteins, DNA and other contaminants. The recovery rates of the expressed FVIII obtained so far were only moderate.
In general and specifically for the recombinant production of proteins (including plasma proteins such as FVIII), it is very important to optimize the process to achieve a high productivity of the product. This is essential for the economy of the product as the recombinant production procedure is relatively expensive and sensitive for disturbances (infections, etc.) due to its biological origin.